October 19, 2017

The Validity of IQ

So what’s the point of IQ tests? Well, they can predict lots of life outcomes. Here is a table from page 65 “The Scientific American Book of the Brain”. It shows the percentages of each IQ bracket and their life outcomes.

IQ is a better predictor of job performance than lots of things. From Schmidt and Hunter, they looked at the correlation between job performance ratings given by coworkers with various metrics to figure out what best predicts subjectively assessed work performance. The best prediction was a work sample test, followed by a structured interview and IQ test:

They also did a review of job performance studies :

Tarmo Strenze reviewed a series of longitudinal studies that compared how various life factors correlated with education level, income and occupation. The average sample size for each group is 97,083, which is the largest average sample size in any psychological study I have ever seen.

In all three groups, IQ was a better predictor for education level, occupation level, and income than any other factor.

Similarly, In the paper “Intelligence and School Grades: A meta-analysis”, Bettina Roth and co  did a meta-analysis of data from more than 200 samples totaling 105,185 students, and shows that IQ tests strongly correlated with grades at 0.54.

So even if you are not convinced that IQ measures “intelligence”, it obviously measures something, and this thing is of practical importance. That said, there are some reasons to believe that IQ tests measure intelligence.

IQ Probably Measures “Intelligence”

We know that IQ tests measure intelligence because IQ tests correlate with peer and self rated intelligence. For instance, in Denissen et al. 2011 489 college students were divided into 20 groups which studied together for a period of one year. At the end of this year, subjects were asked to rate how intelligent their group mates were on a 7 point scale ranging from “not intelligence” to “very intelligent”. It was found that the better a subject did on an IQ test the smarter their group mates thought they were.

Palhusand Morgan 1997 found similar results and also showed that the correlation between peer rated intelligence and IQ increased the longer the peer knew the person being tested. They had 5 group discussion sections, and in the first section, found that in the first session, intelligence ratings were almost entirely a function of how much people talked. By the Fifth session, it was almost entirely a function of the person’s IQ – so talking a lot only increases perceived “intelligence” above IQ in the short term.

Similarly,Bailey and Hatch 1979 showed that intelligence rated by people’s close friends correlated with their IQ and Bailey and Mattetal 1977 found the same was true of spouses.

A significant body of research has also shown that IQ tests predict how intelligent people rate themselves as being (Paulhus, Lysy, and Yik 1998, ,Angelo and James 1977, and Reilly and Mulhern 1995). Clearly then, the smarter a person thinks they are, and the smarter their friends think they are, the better they tend to do on IQ tests.

So in summary, IQ tests predict life outcomes better than several factors commonly recognized to predict life outcomes, such as what your parents are like and how good your grades are. And IQ predicts one’s subjective perception of a person’s intelligence the longer you interact with them.

Facebook Comments
  • Emil Kirkegaard

    There’s a meta-analysis of self-estimated IQ and actual IQ. Mean r = .33. So, adjust for measurement error and it will be higher, perhaps .40 or so. https://www.ncbi.nlm.nih.gov/pubmed/22181852

  • Andreas Egeland

    Although I am somewhat perplexed that a ~.5 correlation is considered sufficient, however I do agree that IQ does measure something, and that this thing is of practical importance.

    Where it all goes wrong is when you suggest IQ then measures intelligence.
    The flaw in your reasoning is pretty clear. You use another metric to define what intelligence really is (self-rating is mentioned, but obviously peer-rated intelligence is a much better metric) and then compare the IQ scores with this. This is only a valid argument if you genuinely believe that peer-rated intelligence is the objectively true measure of intelligence. In reality this is but one of many inexact metrics, so, let me be clear, this only means that IQ is correlated with intelligence. They say correlation does not imply causation, but it is perhaps less well known that a correlating factor is not a deciding factor. IQ thus gives us an indication of someone’s intelligence, not a measurement thereof.

    I do enjoy writing very long comments, so although I now go on to address the studies themselves, I do repeat myself quite a lot (so you might already have read all you need to know about my position).
    What is most immediately obvious is that the results are trivial, when phrased differently. IQ correlates with the what people intuit as intelligence. This says very little about intelligence, but a lot about people, in particular the ones who invented the IQ tests. It is trivial that this is the case. The IQ test would not even be known as an Intelligence Quotient if it did not have anything to do with the common perception of intelligence.

    The very important issue remains: Why does this mean IQ probably measures intelligence? Do you define intelligence to be the sum of all subjective experiences of intelligence? In fact, we now run into the real issue: what is intelligence? I don’t mean to be pedantic (actually, I do), but you must define intelligence for this entire piece to make any sense.
    What you seem to be saying is that IQ scores can predict how you judge other people’s intelligence. So what?
    I do not accept this definition of intelligence.